BIAS IN AGE AND SEX MINNESOTA'S DEER HARVEST REGISTRATION

Transcription

1 This document is made available electronically by the Minnesota Legislative Reference Library as part of an ongoing digital archiving project BIAS IN AGE AND SEX CLASSIFICATIONS FROM MINNESOTA'S DEER HARVEST REGISTRATION David K. Ingebrigtsen and Christopher S. DePerno Minnesota Wildlife Report 13 Minnesota Department of Nataral Raoarces Division of Wildlife Wildlife Popalatioas and Rnearch Unit 2002

2 BIAS IN SEX AND AGE CLASSIFICATIONS FROM MINNESOTA'S DEER HARVEST REGISTRATION DAVID K. INGEBRIGTSEN, Minnesot-i Department of Natural Resources, Wildlife Operations, Box 156, Grand Marais, MN CHRISTOPHER S. DEPERNO, Farmland Wildlife Populations and Research Group, Minnesota Department of Natural Resources, R.R. 1, Box 181, Madelia, MN Abstract: To develop a correction matrix useful for adjusting errors in data collected at mandatory hunter registration stations, white-tailed deer (Odocoileus virginianus) age and sex classification data were collected in southern and western Minnesota during and During , of 1,928 deer registered 285 (14.8%) were incorrectly classified. Results indicated that hunters incorrectly classified white-tailed deer by age and sex: adult males (1.3%), adult females (3.901'), fawn males (25.7%) and fawn females (44.6%). During 1988, of 1,566 deer registered 214 ( 13. 7%) were incorrectly classified. Hunters incorrectly classified adult males (1.S-At), adult femal~ (5.8%), fawn males (38.8%) and fawn females (45.3%). An inverse correction matrix was developed for both sampling periods. To improve the accuracy of classifying the harvest data by age and sex, the correction matrix generated fiom this study was applied to Minnesota's 1984 deer registration data. Harvest data are the primary information used for population modeling and determining harvest season regulations in farmland Minnesota. Additionally, both matrices ( and 1988) were used to correct the 1988 farmland :zone harvest data by age and sex and the results were compared. Correction matrices can be utiliz.ed to improve accuracy and provide inexpensive estimates of age and sex clusifications. Five types of errors that may have contributed to registration errors are discussed. Since 1972, the Minnesota Department of Natural Resources (DNR) has collected white-tailed deer (Odocoileus virginianus) harvest data at nwvjatory deer registration stations. Deer registration stations (e.g., gas stations, convenience stores, and lockers) were operated by private citi7.ens not trained in deer aging and sexing techniques. Hunters detached a stub fiom their hunting license and exchanged it for a transport tag at the registration station. The registration station operator marked a box on the registration stub indicating whether the deer was an adult male, adult female, fawn male, or fawn female. Operators seldom ex.unined deer to verify hunter's identification of age and sex. Mi.on. Wildl. Rep. 13, 2002

3 lngebrigtscn and DePemo Paae2 Karns ( 1974) verified that dctcnnination of age and sex at Minnesota registration stations was subject to error and hypothesized that error rates may vary between regions of the state. Age and sex data from hunter registration must be adjusted for registration bias before it is used for population modeling and determining harvest season regulations. Registration bias can be evaluated on a subsample of the registration data by trained personnel at deer registration stations by simultaneously evaluating age and sex determinations made by hunters. Our objective was to develop a registration error correction factor derived from simultaneous age and sex determinations at hunter registration stations that could be used for population modeling and determining harvest season regulations in farmland Minnesota. This study was conducted with assistance ofdnr Section of Wildlife personnel. We thank J. R. Ludwig and A. H. Berner who were instrumental in devising the survey and D. M. Heisey for his advice on the statistical analysis. We thank A. H. Bemer, R. 0. Kimmel, D. E. Simon, and R. Lake for their comments on earlier drafts of this manuscript. METHODS DNR personnel with experience in deer age and sex identification collected deer teeth and aged and sexed white-tailed deer at 6 mandatory deer registration stations in Big Woods Southeast Deer Management Subunit (DMSU) in 1982, 15 stations in Prairie River, Prairie Southwest, Prairie Southeast, Big Woods Metro, and Big Woods Central DMSUs in 1983, and 16 stations in all DMSUs except Red River West and Big Woods Metro in 1988 (Fig. 1 ). To maximi7.e the sample, registration stations with large numbers of deer registered in previous years were selected for the study. During the study, Minnesota had S deer hunting zones with varied regulations. In 1982, the study was conducted in Zone 3 which had a 9-day buck-only season followed by a 3-day buck or antlerless-deer-by-permit season. In 1983, the study was conducted in Zone 4 which had a 3-day buck-only hunt followed by a 2-day season for buck or antlerless-deer-by-pennit. In 1988, the study was conducted in Zone 3 and Zone 4. Zone 3 now had a 9-day buck-only season followed by a 7-day buck or antlerless-deer-by-permit season. Zone 4 now had a 2-day anderless-deer-by-permit season followed by a 4-day antlerless-deer-by-pennit season. The study was conducted on the first 2 days of the antlerless-decr-by-permit seasons. Upon arrival at deer registration stations, hunters registering deer were asked if DNR personnel could remove the deer's front incisors for aging; in 1982 and 1983 a shoulder patch was offered to the hunter as an incentive. The hunter was then instructed to register deer with the rqistration station operator. The registration station operator registered deer inside the station by collecting the hunter's license stub, recording deer age and sex (usually relying on hunter's classification), if necessary inspecting the antlerless pennit, and prescntin& hunter with a transport taa to be affixed to the deer. Operators rarely inspected deer.

4 lngcbrigtsen and DePemo Page 3 Owing the time the hunter was completing registration inside the statio~ a DNR employee remained outside the registration station and extracted the front incisors, determined sex by appearance of antler growth and/or genitalia, and distinguished fawns from adults by tooth replacement and wear (Severinghaus, 1949). The DNR employee discreetly recorded the above information in addition to the hunter's license number which was printed on the hunting license, possession tag, and registration stub. At close of registration each day, Minnesota DNR Section of Wildlife personnel inspected registration stubs and recorded the registration station operator's data. Therefore, two chwifications were collected for each deer: 1) age and sex determination by the nonnal registration procedure, and 2) age and sex detennination conducted by DNR personnel. To insure that method of sampling would have minimal bias, DNR personnel were instructed to act casually and not give the impression they were evaluating the registration procedure. Registration station operators and hunters were instructed to register deer by their normal procedures. Prior to registering their deer, hunters were not infonned of their deer's age or sex. Furthermore, to prevent hunters and registration station operators from being aware the registration procedure was being evaluated, all information was discreetly recorded by the DNR employee. One method often used to analyze data and apply correction factors assumes that populations sampled have identical age and sex compositions. For example, given the matrix derived from concurmit sex/age classifications, and given a new set of error-prone data (S., T., U V.) observed by registration station harvest classification, the estimate of adult Rclristration Stub Classification 'True' Class:.1cation 1 Age/Sex Adult Adult Fawn Fawn Male Female Male Female Adult Male a b c d w Adult Female e f g h x Fawn Male J k l y Fawn Female m n 0 p z Registration s T u v Total 'Age and sex classification determined by Minnesota DNR personnel. males adjusted by the registration bias matrix W 1111 would be calculated from proportions of rqistration totals that were correctly registered in this study.

5 lngebrigtsen and DePemo Page4 However. the above is the sum of proportions of actual age and sex classifications in the subsample multiplied by the observed values. Therefore, the adjusted value depends on he proportion of age and sex class in the sample (e.g., a/s is the proportion of registered adult males that were actually adult males, thus [a/s]*s depends on the proportion of adult males in the sample). This method is only useful for adjusting a sample of registration data with similar proportions of actual sex/age classifications as the subsample used to create correction factors. However, actual sex and age proportions usually differ in harvested populations. MacDonald and Dillman ( 1968: 123) compared a mail survey of deer hunters with a parallel data set of their known perf onnance and noted that interpretation of data depended upon the actual relative proportion of age and sex classes harvested. A more precise method of analysis is i;cquired for adjusting data of other populations. This can be accomplished by deriving a correction matrix (Hoenig and Heisey 1987) from analysis of proportions of "true" (i.e., detennined by DNR biologist.s) age and sex classifications that were correctly registered by hunters. Thus, a new set of error-prone deer registration data can be adjusted by an inverse matrix developed from this study, regardless of sex/age class proportions. To determine the correction matrix needed to transpose errors, data were classified and entered in cells. From analysis of proportions ofdnr-determined age and sex classifications that were incorrectly registered (Table 1 ), a correction matrix (Table 2) was generated by inverting the matrix (Hoenig and Heisey 1987). This technique can be used to adjust individual registration data to obtain harvest estimates, which are essential for accurate population modeling efforts. Data analysis was conducted with a computer spreadsheet. A chi-square goodness of fit test was used to test if data from DMSUs could be pooled. RESULTS AND DISCUSSION Concmrcnt age and sex classifications were obtained for 638 deer at 6 selected registration stations in 1982 and 1,290 deer at 15 selected registration stations in No difference between error classification matrices was noted among the 6 DMSUs (X , df = 60, f = ). Therefore, for pooling of DMSUs is a reasonable method of developing a correction matrix useful for adjusting fannland deer harvest estimates (Tables l, 2). Concurrent age and sex classifications were obtained for 1,566 deer at 16 registration stations in No difference between error classification matrices was found amona the 9 DMSUs (X 2 = 99.l, df = 96, f = 0.394). Therefore, for 1988 pooling of DMSUs is a reasonable method of developing a correction matrix useful for ad,,.jsting farmland deer harvest estimates (Tables 3, 4).

6 lngebrigtsen and DePemo f>aaes During , hunters incorrectly registered 285 (14.8%) deer. Female fawns were incorrectly registered most often (44.6%) followed by male fawns (25.7%). Very few errors were made in registering adult males (l.3%) or adult females (3.90/o). Hunter registration underestimated number of fawns harvested by 24.5% and overestimated number of adult females by 22.2%. Similarly in 1988, hunters incorrectly registered 214 (13.7%) deer. Female fawns were incorrectly registered most often (45.3%) followed by male fawns (38.8%). Very few errors were made in registering adult males (1.8%) or adult females (5.8%). During 1988, hunters underestimated number of fawns harvested by /o and overestimated number of adult females by 18.8%. Our observations suggest that registration errors most likely resulted from: 1. Hunter erred or falsified the sex/age identification of harvested white-tailed deer. This may be due to peer pressure. Many hunters feel embarrassed by shooting a young deer and may register their deer as an adult to avoid admitting they shot a fawn. 2. Antlerless deer were incorrectly classified as a "doe" by hunters and thereby registered as females because antlerless permits are incorrectly krf>wn as 1 '1N' pe.. mits 1 rather than 'either sex permits. 3. In cases where the registration operator inspects deer, the operator erred or falsified sex/age identification of harvested white-tailed deer. 4. Hunters tagged a deer shot by another member of the hunting party. In this situation the hunter may not be aware of what another member shot and what he actually tagged. 5. Random recording errors by registration station operators or DNR personnel. In southern and western Minnesota, over 9()0/o of adult males have antlers with 4 or more points. If many of the errors were truly random, some adult bucks would have been classified as female fawns. When combining periods and 1988, only one adult male was identified on registration stubs as a female fawn (Tables 1, 3), thus few mistakes were probably made as a result of random errors. The inverse correction matrix generated from pooled study data (Table 2) was used to adjust the tally of 62,265 deer registered during the 1984 antlerless season in farmland Minnesota (Table 5). To apply the correction matrix, registration stub data totals wete multiplied vertically within the matri.'t and cells added horizontally to produce corrected classifications of sex and age. Results of applying the correction matrix to the 1984 farmland deer registration data indicates that hunters under-reported the number of fawns harvested by 23.8%, over-reported nwnber of adult females by 18.90/o, and over-reported the proportion of males (55.1%) in the fawn harvest (adjusted proportion= 51.2%; Table 5). Within individual registration blocks, the magnitude of difference among registration data and adjusted figures was much greater.

7 lngebriatscn and DePemo Page6 Because misclassification of fawns was a common classification error, registration errors varied with percentage of fawns in the harvest. Similarly, other studies have documented a large error rate of fawn identification. In northern Minncso~ Karns (1974) showed that deer hunter registration data underestimated fawn numbers by 24%, whereas in Nebraska, Trindle and Menzel (1985) demonstrated that hunter registration underestimated fawn numbers by 64%. Furthermore, hunter surveys compared with professional age and sex detenninations at check stations noted that white-tailed deer fawns were underestimated by 54% (Smith 1959) and 26% (MacDonald and Dillman 1968). Application of the 1988 error classification matrix (Table 4) to 1988 fannland registration data corrected the harvest totals of fawn males and fawn females by 41.00/e and 1.3%, respectively (Table 6). Interestingly, the matrix corrected the 1988 harvest totals for fawn males by and fawn females by 39.8% and was significantly different than the correction by the 1988 matrix (X 2 = 32.3, df= 12, f <0.001). In 1988, hunters made more mistakes when registering male fawns. From the above matrix corrections it appears that more fawn males were incorrectly classified in 1988, whereas more fawn females were incorrectly classified in (Table 6). However, the ratio of male fawns to female fawns registered (i.e., in the harvest) did not change substantially from (57.8%:42.2%; Table 1) to 1988 (52.2%:47.8%; Table 3 ). Adult males made up 42. 7% of the deer registered in 1988 (Table 3) and 26.,.;. of the deer registered in (Table 1), whereas adult females made up 27.9% of the deer registered in 1988 (Table 3) and 33.5% of the deer registered in (Table 1 ). These comparisons suggest the source of primary error in hunter registration may not be an error in sex identification but an error of including fawns into the wrong age class. Certain biases arc inherent in our study. Because of DNR presence at registration stations, hunters may have been less inclined to provide incorrect age and sex information or the station operator may have altered his inspection routine. These biases were probably minimal compared to the magnitude of registration error and may have been partially self-canceling. This study was designed to minimize bias by not revealing the full intent to registration operators, inspecting registration stubs once at the end of each day, and informing operators the intent of stub inspection was only to count number of deer registered. The 1988 correction matrix (Table 4) is only applicable to buck or antlcrless-decr-by-pennit seasons in fannland Minnesota. Changes in season format and/or hunter and registration station operator knowledge may require further study and the development of a new matrix. In other regions where spike antlers arc common on adult deer, more errors in aging deer may occur.

8 Inaebriatscn and DcPcmo Pqc7 MANAGEMENT RECOMMENDATIONS Harvest data collection should be continued at mandatory registration stations and the correction matrix should be applied to adjust for registration bias. This technique provides a reasonable estimate of registration classifications and is the most cost-efficient method of acquiring this data. The registration cojtection matrix from this study should be applied to all buck or antlcrlcss-dccr-by-pennit season harvest registration figures in farmland Minnesota for purposes of population modeling and determining harvest regulations. The study should be repeated at least every 10 years or whenever the hwtting season format changes. LITERATURE CITED Hoenig, J. M. and D. M. Heisey Use of a log-linear model with the EM algorithm to correct estimates of stock composition and to convert length to age. Trans. Amer. Fish. Soc. 116: Karns, P. D Hunter identification of deer age and sex. Minn. Wildt. Quar. 34:1-5. MacDonald, D. and E. G. Dillman Techniques for estimating non-statistical bias in big game harvest surveys. J. Wildl. Manage. 32: Severingbaus, C. W Tooth development and wear as criteria of age in white-tailed deer. J. W'tldl. Manage. 13: Smith R. H A hunter questionnaire for compiling hwtt statistics. Proc. Annu. Conf. West. Assoc. State Game and Fish Comm. 39: Trindle, B. and K. Menzel Surveys and management of deer. Nebraska Game and Parks Comm. PR Proj. W-15-R pp.

9 lnpbriatsen and DcPemo Pqc8 Table 1. Aac and sex classifications from selected farmland deer registration stations, Minnesota (Correct age and sex registrations in parentheses). Registration Stub Classification 'True' Classification Age/Sex Adult Adult Fawn Fawn Male Female Male Female Adult Male (514) Adult Female 4 (645) Fawn Male (300) Fawn Female (184) 332 Registration Total ,928 Age and sex clusification determined by Minnesota DNR personnel.

10 lngebrigtscn and DePemo Page9 Table 2. Inverse matrix of age and sex clmsitications from selected farmland deer registration stations, Minnesota Cell values mwtiplied by respective columns registration total and summed is equal to the true classification. Registration Stub Classification 'True' Classification Age/Sex Adult Adult Fawn Fawn Male Female Male Female Adult Male Adult Female Fawn Male Fawn Female Registration Total 'Age and sex classification dctennined by Minnesota DNR personnel.

11 Inpbriatsen and ~Perno Pqc 10 Table 3. Aae and sex classifications from selected farmland deer registration stations, Minnesota (Correct age and sex registrations in paienthcses). Registration Stub Classification 'True' Classification Age/Sex Adult Adult Fawn Fawn Male Female Male Female Adult Male (669) Adult Female 3 (437) Fawn Male (147) Fawn Female (99) 181 Registration Total ,566 Age and sex classification determined by Minnesota DNR personnel.

12 Inaebriauen and DePcmo Page 12 Table 5. Farmland Minnesota's 1984 deer harvest adjusted by age and sex usin& the inverse matrix of data collected al 21 farmland deer registration stations, Minnesota Cell values multiplied by respective columns registration total and summed is equal to the 1n1c, classification. Registration Stub Classification 'True Classification Adult Adult Fawn Fawn Male Female Male Female Adult Male ,167 Adult Female n Fawn Male ,514 Fawn Female ,112 ~Total 27,633 21,955 6,984 S,693 62,265 Aae md sex cbmificmion determined by Minnesota DNR penonncl.

13 lngebrigtsen and DePemo J>aee 13 Table 6. Minnesota's 1988 farm.land deer harvest data and the deer harvest data corrected by age and sex using the matrix and the 1988 matrix. Adult Male Adult Female Fawn Male Fawn Female 1911 Harvest dala (raw) 36,903 24,834 7,608 5, Comded Harvest dmausin&: 1911 Matrix 36,808 21,730 10,729 6,001 (%cbmae ) (-0.3%) (-12.5%) (41.0%) (1.3%) Matrix 36,419 21, ,213 (%cbmae ) (-lj!.{ ) (-14.3%) (21.0%) (39.8%) Conectecl from raw dma.

14 Inaebrigtsen and DePcmo Paac 14 Figure 1. Deer management subwiits of farmland Minnesota, and 1988.

15 ?vfinnesota WILDLIFE REPORTS 1. Attracting broods with tape-recorded chick calls. Richard 0. Kimmel pp. 2. Foods and feeding behavior of young ~y partridge in Minnesota. Ricblrd Erpelding, Richard 0. Kimmel, and Drake J. Lockman pp. 3. EfJC\.;ivcncss ofswarcflex wildlife warning reflectors in reducing deer-vehicle collisions in Minnesota. David K. lngebritscn and John R. Ludwig pp. 4. Effects of hunting seasons on pheasant populations. Alfred H. Bemer and Dennis Heisey pp. S. Mcasurancnt error models for commo1. wildlife radio-tnckiog systems. Richard M. Pace, ill pp. 6. Hunting harvest of wbifc..tailed deer in the Bearville Study Area, nortbcenlral Minne.410ta. Todd K. Fuller pp. 7. Summary and evaluation of Minnesota's waterfowl breeding populltion survey, 19n Stephen J. Maxson and Richard M. Pace, m pp. 8. An annotated bibliography of forest management for white-tailed deer. Mart S. Lenarz pp. 9. The white-tailed deer in Minnesota's Farmland: the 1960's. Alfied H. Bemer llld Damis Simon (compilers) pp. 10. The sharp-tailed grouse in Minnesota. William E. Berg l 7pp. 11. The greater prairie-chicken in Minnesota. W. Daniel Svedmsky, Tenance J. Wolfe, andjobne. Toepfer pp. 12. Assessment of a mallard model in Minnesota's Prairie Coceau. MicMel C. Zicus and David P. Rave pp. 13. Bias of sex and age classifications from Minnesota's deeer harvest registralion. David K. lngebrigtxn and Christopher S. DePcmo pp.